Recent technological advances have made it possible to use three-dimensional MEMS, micro-optical devices and other micro-structures in a variety of fields, including photonics, communications, and integrated circuits. In the past, these tiny devices were fabricated using laser micro-machining tools. However, this method was time consuming and expensive, and thus, it was typically difficult for manufacturers to meet production requirements in a cost efficient manner. In this regard, such techniques did not work well with commonly applied techniques for manufacturing integrated circuits because each pixel of the design had to be rewritten using a new algorithm. Since this was a laborious and time-consuming undertaking, many have avoided the use of micro-machining tools.
In light of the desirability to use small scale, three-dimensional structures, other manufacturing techniques have been developed in an attempt to avoid the problems associated with laser micro-machining tools. In particular, traditional optical lithography techniques used for fabricating integrated circuits have been adapted to manufacture three-dimensional microstructures. In traditional optical lithography, a fully resolved pattern is etched into a binary photomask and transferred to a wafer by exposing the wafer through an exposure tool (e.g., stepper). More particularly, binary photomasks are typically comprised of a substantially transparent substrate (e.g., quartz) and an opaque layer (e.g., chrome) in which the pattern to be transferred is etched. It is also known that other layers may be included on the photomask, including, for example, an antireflective layer (e.g., chrome oxide). The photoresist in the substrate on the integrated circuit being processed is then developed and either the exposed or unexposed portions are removed. Thereafter, the material on the substrate is etched in the areas where the photoresist is removed. An example of the technology involved in manufacturing a traditional binary photomask (e.g., chrome-on-quartz) and its use to manufacture integrated circuits is disclosed in, for example, U.S. Pat. No. 6,406,818.
These known processes for fabricating binary photomasks and semiconductor devices have been modified for the manufacture of three-dimensional, microscopic devices. In this regard, it is known to use a continuous tone pattern on the photomask (e.g., chrome-on-glass) instead of a binary, fully resolved mask pattern to yield a continuous tone intensity through the photomask during image formation. One type of continuous tone, variable transmission photomask is commonly known as a binary half tone (“BHT”) photomask. BHT photomasks use two levels of gray tones (e.g., 0% transmissive and 100% transmissive). Another type of continuous tone, variable transmission photomask is known as gray scale photomasks, which use varying levels of transmission of light through the photomask (e.g., 0%, 50%, 100%, etc.). By using these types of variable transmission photomasks, a three-dimensional structure can be formed in the photoresist on a wafer through the use of a continuous tone pattern.
BHT photomasks are typically designed to have sub-resolution features that partially transmit exposure source light intensity based on feature modulation in width and pitch. In this regard, it is known in the art to design a BHT photomask layout for microscopic surfaces by dividing the patterned area of the photomask into pixels and sub-pixels (commonly referred to as “sub-pixelation”) which define areas on the mask through which light is to be transmitted, as shown in FIGS. 1 and 2. The sub-pixels defining the BHT photomask pattern are designed to be smaller than the resolution of the exposure tool being used so that a gray scale image can be created on the resulting wafer. The boundaries of the sub-pixel's size are typically defined by Rayleigh's equation (1) as follows:R=kÿ/NA   (1)where R is the minimum resolvable half pitch feature of the wafer, ÿ is the exposure tool wavelength, NA is the numerical aperture of the optical system of the exposure tool being used and k (k factor) is a unitless constant whose value depends on process capability (e.g., the smaller the k factor, the better low contrast aerial images can be seen). Generally speaking, the sub-pixels that are required by gray scale designs need to be unresolved in the imaging system, and thus, the k factor should preferably be less than 0.5. As a practical matter, however, the k factor can be somewhat greater than 0.5 and still be unresolved by the total process for some exposure tools. Photomask designers have used calibrated simulation tools, such as the Prolith/2 manufactured by KLA-Tencor, to converge on the optimum unresolvable feature size. Unfortunately, there are many other tools which do not meet the requirements of equation (1), and thus, the design of the photomask is often limited by the capabilities of the design tools available to the designer. Moreover, since photomasks are commonly designed to include other structures (e.g., two dimensional, binary components such as integrated circuit patterns) in addition to a three-dimensional device (e.g., photonics application), this problem is even more complex than implied by the above equation. In such cases, certain BHT cell designs may involve isolated spaces in chrome or chrome islands on the mask and the notion of half pitch is not defined.
As understood by those skilled in the art, tolerable surface roughness effects the minimum feature size in the device under fabrication. For example, where the k factor is 0.7 in equation (1) (i.e., the minimum feature size is resolvable by the optical system), an attempt to construct a BHT photomask having a step-ramp layout will result in a sub-ripple within each step of the ramp pattern, as shown in FIG. 3. In some applications where the specifications of the device permit, a ripple-effect may be acceptable, albeit undesirable. However, in many applications, a sub-ripple effect is not acceptable since a smoother profile is needed for optimal performance of the device being fabricated. Since the prior art BHT photomask design requires the use of a small k factor to achieve a smoother profile, the mask designer is limited to equipment meeting this requirement.
In addition to the k factor, the design of a BHT photomask layout is governed by other specifications of the optical lithography equipment being used, including, for example, its resolution, magnification, wavelength, etc. In this regard, known gray scale applications, such as micro-optical surface generation, require data to have a higher resolution than what is typical of most mask pattern generators. As a result, the mask designer is limited to only those write tools that have the ability to match the gray level grid design associated therewith. For example, an electron beam write tool such as the MEBES 4500, with a write grid of 20 nm cannot properly replicate a BHT design whose sub-pixel variations is 10 nm. A laser beam write tool, by contrast, such as the ALTA 3500 having a write address of 5 nm is capable of replicating the same design.
Moreover, an imaging solution requiring custom materials at the mask will often add cost and complexity in the overall manufacturing process due to the difficulties typically associated with integrating new materials into a photomask. For example, it is known in the art to use variable attenuating films (“VAF”), rather than a BHT photomask, to make three-dimensional devices. However, VAFs are typically expensive and yield less than desirable results.
Once the mask pattern design is completed, the design is transferred to the photomask using optical lithography methods similar to those used to process a conventional binary photomask, as shown in FIG. 4. More particularly, a binary photomask having photoresist 51, chrome 53 and quartz 55 layers is placed under a photomask pattern generator. The photoresist layer 51 is exposed to an optical, laser, electron beam or other write tool in accordance with the data file defining the BHT photomask. The exposed portions of the photoresist layer 51 are developed (i.e., removed) to expose the underlying chrome portions of the chrome layer 53. Next, the exposed chrome portions are etched away (e.g., by dry plasma or wet etching techniques). Thereafter, the remaining photoresist 51 is removed to form a completed BHT photomask in accordance with the BHT photomask layout.
The variable intensity gray tone pattern defined by the mask is next transferred to a wafer coated with photoresist using a wafer stepper or other optical lithography tools. More specifically, varying light intensities are exposed to portions of photoresist on the wafer as defined by the openings in the BHT photomask. The photoresist, in turn, exhibits changes in optical density and a gray scale profile is created thereon. It is noted, however, that the photoresist process is often limited to the variable dose pattern generator being used. Next, the exposed photoresist is removed and the remaining photoresist forms a gray scale pattern which corresponds to the mask design. The photoresist and wafer are then etched to predetermined depths to conform to the gray scale pattern. The result is a three-dimensional micro-structure on the wafer.
In the known methods described above, the minimum feature size suiting the needed application (e.g., three-dimensional microscopic structure) is determined through known simulation techniques, by experimentation or other methods. Once the minimum feature size is determined, a pixel defining the mask dimensions is generated. Various methods have been applied to array a gray scale design (e.g., squares, pixels or spots using variable pitch and variable sub-pixelation methods). These methods, however, have their limitations. In this regard, it is known that contact hole and spot features are more difficult to control than line and space features. As a result, both corner rounding and linearity of the design are compromised. Similarly, variable pitch methods are problematic since they require a different algorithm to be applied at each pixel position to carefully define the correct opening in the BHT mask. When considering the dynamic range for the layout, the square pixel changes non-linearly since the size is varied over the contact area. This can limit the ability of a mask designer to make the fine changes required to correct for process non-linearity. These methods add to the costs and processing time in preparation of the mask since they require a large number of adjustments to be made to the overall design.
One example of a known method for designing a BHT photomask is described in U.S. Pat. No. 5,310,623 (the ‘623 Patent’). The '623 Patent teaches a method for fabricating microlenses through the use of a single exposure mask with precisely located and sized light transmitting openings to enable an image replica to be produced in photoresist materials, and ultimately transferred to a substrate. As disclosed, the design for the photomask is generated using three-dimensional modeling software, wherein a single pixel defines the shape of the microlens. The single pixel is sub-divided into “sub-pixels”, which in turn are sub-divided into gray scale resolution elements. Each sub-pixel and gray scale resolution elements is designed to be “equal [in] size” on each side. (Col. 6, line 30). In other words, each sub-pixel and gray scale element is a perfect square.
U.S. Pat. No. 6,335,151 discloses a method for fabricating a microscopic, three-dimensional object by creating a mask having consisting of pixels and “super-pixels” which define the contours of the object's surface, imaging the mask's pattern onto a photoresist film, and transferring the three-dimensional surface from the photoresist to a substrate.
Although useful, the conventional square pixel array methods used in the prior art have their shortcomings. In this regard, the prior art discloses the use of square pixels having a size which is less than the minimum resolvable feature size of an optical system. Each pixel is then divided into sub-pixels whose respective areas are changed in both the x and y axes, as disclosed in the '151 and '623 Patents. As a result, the change in area of each sub-pixel is a square of the amplitude, thereby making the change in light intensity between each gray level non-linear and often infinite. Thus, the transmission of light is limited by the square sub-pixel's size as well as the minimum dimensions permitted by the mask pattern generator. Accordingly, three-dimensional objects made by the prior art methods tend to have jagged surfaces, especially where the objects are sloped. Since these methods produce marginal results, many have migrated away from the use of BHT photomasks for making three-dimensional microscopic devices.
Thus, there is a long felt need for new design rules and layout choices for making BHT photomasks to overcome these shortcoming associated with the prior art.
While the prior art is of interest, the known methods and apparatus of the prior art present several limitations which the present invention seeks to overcome.
In particular, it is an object of the present invention to provide a method for designing a BHT photomask layout having a smooth profile.
It is a further object of the present invention to provide a method for designing a BHT photomask layout which meets the specifications of a wide range of optical systems.
It is another object of the present invention to design a BHT photomask wherein the change in light intensity between each gray level is both linear and finite.
It is another object of the present invention to solve the shortcomings of the prior art.
Other objects will become apparent from the foregoing description.